zlh1992

老师 目前对于第一课代码有个疑问。 #将label变成one-hot编码，因为softmax_propabilities是一个数组，是10个概率，每个概率代表着预测结果属于其index类的概率，为了计算交叉熵，我们需要把label也转换成一个数组 self.label_tensor=tf.one_hot(self.label_tensor,10) #计算交叉熵 cross_entropy=tf.nn.softmax_cross_entropy_with_logits(logits=self.fc_result,labels=self.label_tensor) 您修改成这个样子了 目前跑不通 报错： ValueError: Cannot feed value of shape (1000,) for Tensor 'one_hot:0', which has shape '(1000, 10)' 去掉onehot 仍然修改为cross_entropy=tf.nn.sparse_softmax_cross_entropy_with_logits(logits=self.fc_result,labels=self.label_tensor) 可以跑通,但是模型不收敛。acc一直很低。

并没有提供训练代码。。

> do you solve the problem? I just meet the same problem. the loss of eval_data_set is very low, but I got the same result, when I use the alpaca...

/opt/conda/envs/tch/lib/python3.9/site-packages/peft/tuners/lora.py:619 in forward │ │ │ │ 616 │ │ │ │ self.unmerge() │ │ 617 │ │ │ result = F.linear(x, transpose(self.weight, self.fan_in_fan_out), bias=self. │ │ 618 │ │...

从显存占用看没有比paged_8bit_lion少..

同求

how to set lora arguments when I have multi GPUs?

I just meet same problem. Did you solved it?

我在pycharm里自己起了mysql服务 可以不报错 在 2024-06-13 15:36:11，"姜永久" ***@***.***> 写道： 我也遇到一样的问题，需要起MySQL服务？ — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you authored the thread.Message ID:...

> @chuangzhidan Did you check out and install vllm from this PR? You seem to have a different version: > > ``` > root@8b74d742fc51:~/vllm# pip show vllm > Name: vllm...